Introduction to geopandas and cartopy

Basic Setup

Again we will be using pandas and matplotlib in this tutorial.

import pandas as pd
import matplotlib.pyplot as plt

We’ll also supress a few disturbing warnings.

import warnings
warnings.filterwarnings('ignore')

Why do we need something other than pandas?

Let’s reload again our example dataset of conventional power plants in Europe as a pd.DataFrame.

fn = "https://raw.githubusercontent.com/PyPSA/powerplantmatching/master/powerplants.csv"
ppl = pd.read_csv(fn, index_col=0)

This dataset includes coordinates (latitude and longitude), which allows us to plot the location and capacity of all power plants in a scatter plot:

ppl.plot.scatter('lon', 'lat', s=ppl.Capacity/1e3)

<Axes: xlabel='lon', ylabel='lat'>

However, this graphs misses some geographic reference point, we’d normally expect for a map like shorelines, country borders etc.

Geopandas - a Pandas extension for geospatial data

Geopandas extends pandas by adding support for geospatial data.

The core data structure in GeoPandas is the geopandas.GeoDataFrame, a subclass of pandas.DataFrame, that can store geometry columns and perform spatial operations.

Note

Documentation for this package is available at https://geopandas.org/en/stable/.

Typical geometries are points, lines, and polygons. They come from another library called shapely.

First, we need to import the geopandas package. The conventional alias is gpd:

import geopandas as gpd

We can convert the latitude and longitude values given in the dataset to formal geometries (to be exact: shapely.Point objects but we won’t go into detail regarding this) using the gpd.points_from_xy() function, and use this to gpd.GeoDataFrame. We should also specify a so-called coordinate reference system (CRS). The code ‘4326’ means latitude and longitude values.

geometry = gpd.points_from_xy(ppl['lon'], ppl['lat'])
gdf = gpd.GeoDataFrame(ppl, geometry=geometry, crs=4326)

Now, the gdf looks like this:

gdf.head(3)

	Name	Fueltype	Technology	Set	Country	Capacity	Efficiency	DateIn	DateRetrofit	DateOut	lat	lon	Duration	Volume_Mm3	DamHeight_m	StorageCapacity_MWh	EIC	projectID	geometry
id
0	Emsland	Nuclear	Steam Turbine	CHP	Germany	1336.000000	0.33	1988.0	1988.0	2022.0	52.481878	7.306658	NaN	0.0	0.0	0.0	{'11WD7KKE-1K--KW5'}	{'ENTSOE': {'11WD7KKE-1K--KW5'}, 'OPSD': {'BNA...	POINT (7.30666 52.48188)
1	Mellach	Hard Coal	Steam Turbine	CHP	Austria	200.000000	NaN	1986.0	1986.0	2020.0	46.911700	15.488300	NaN	0.0	0.0	0.0	{'14W-WML-KW-----0'}	{'BEYONDCOAL': {'BEYOND-AT-11'}, 'ENTSOE': {'1...	POINT (15.48830 46.91170)
2	Eemshaven	Hard Coal	CCGT	PP	Netherlands	1604.170304	0.58	2015.0	NaN	2029.0	53.440500	6.861200	NaN	0.0	0.0	0.0	{'49W000000000044-'}	{'BEYONDCOAL': {'BEYOND-NL-12'}, 'ENTSOE': {'4...	POINT (6.86120 53.44050)

With the additional geometry columns, it is now even easier to plot the geographic data:

gdf.plot(
    column='Fueltype',
    markersize=gdf.Capacity/1e2,
)

<Axes: >

We can also start up an interactive map to explore the geodata in more detail:

gdf.explore(column='Fueltype')

Make this Notebook Trusted to load map: File -> Trust Notebook

Map Projections with Cartopy

Cartopy is a Python package designed for geospatial data processing and has exposed an interface to enable easy map creation using matplotlib.

The Earth is a globe, but we present maps usually on two-dimensional surfaces. Hence, we typically need to project data points onto flat surfaces (e.g. screens, paper). However, we will always loose some information in doing so.

A map projection is:

a systematic transformation of the latitudes and longitudes of locations from the surface of a sphere or an ellipsoid into locations on a plane. Wikipedia: Map projection.

Different projections preserve different metric properties. As a result, converting geodata from one projection to another is a common exercise in geographic data science.

• conformal projections preserve angles/directions (e.g. Mercator projection)

• equal-area projections preserve area measure (e.g. Mollweide)

• equidistant projections preserve distances between points (e.g. Plate carrée)

• compromise projections seek to strike a balance between distortions (e.g. Robinson)

If you like the “Orange-as-Earth” analogy for projections, checkout this numberphile video by Hannah Fry.

Note

Documentation for this package is available at https://scitools.org.uk/cartopy/docs/latest/.

First, we need to import the relevant parts of the cartopy package:

import cartopy
import cartopy.crs as ccrs

Let’s draw a first map with cartopy outlining the global coastlines in the so-called plate carrée projection (equirectangular projection):

ax = plt.axes(projection=ccrs.PlateCarree())
ax.coastlines()

<cartopy.mpl.feature_artist.FeatureArtist at 0x7f1954c6eb10>

A list of the available projections can be found on the Cartopy projection list page.

ax = plt.axes(projection=ccrs.Mollweide())
ax.stock_img()

<matplotlib.image.AxesImage at 0x7f1955f0e950>

We can combine the functionality of cartopy with geopandas plots:

fig = plt.figure(figsize=(7,7))

ax = plt.axes(projection=ccrs.PlateCarree())

gdf.plot(
    ax=ax,
    column='Fueltype',
    markersize=gdf.Capacity/1e2,
)

<GeoAxes: >

We can add further geographic features to this map for better orientation.

For instance, we can add the coastlines…

ax.coastlines()
fig

… country borders …

ax.add_feature(cartopy.feature.BORDERS, color='grey', linewidth=0.5)
fig

… colour in the ocean in blue …

ax.add_feature(cartopy.feature.OCEAN, color='azure')
fig

…and color in the land area in yellow …

ax.add_feature(cartopy.feature.LAND, color='cornsilk')
fig

Geopandas will automatically calculate sensible bounds for the plot given the geographic data. But we can also manually zoom in or out by setting the spatial extent with the .set_extent() method:

ax.set_extent([5, 16, 47, 55])
fig

Reprojecting a GeoDataFrame

In geopandas, we can use the function .to_crs() to convert a GeoDataFrame to a desired coordinate reference system. In this particular case, we use the proj4_init string of an initialised cartopy projection to reproject our power plant GeoDataFrame.

fig = plt.figure(figsize=(7,7))

crs = ccrs.AlbersEqualArea()

ax = plt.axes(projection=crs)

gdf.to_crs(crs.proj4_init).plot(
    ax=ax,
    column='Fueltype',
    markersize=gdf.Capacity/1e2,
)

ax.coastlines()

<cartopy.mpl.feature_artist.FeatureArtist at 0x7f195545bb50>

Reading and Writing Files with geopandas

In the following example, we’ll load a dataset containing the NUTS regions:

Nomenclature of Territorial Units for Statistics or NUTS (French: Nomenclature des unités territoriales statistiques) is a geocode standard for referencing the subdivisions of countries for statistical purposes.

Our ultimate goal for this part of the tutorial is to map the power plant capacities to the NUTS-1 region they belong to.

Common filetypes for vector-based geospatial datasets are GeoPackage (.gpkg), GeoJSON (.geojson), File Geodatabase (.gdb), or Shapefiles (.shp).

In geopandas we can use the gpd.read_file() function to read such files. So let’s start:

url = "https://tubcloud.tu-berlin.de/s/RHZJrN8Dnfn26nr/download/NUTS_RG_10M_2021_4326.geojson"

nuts = gpd.read_file(url)

nuts.head(3)

	id	NUTS_ID	LEVL_CODE	CNTR_CODE	NAME_LATN	NUTS_NAME	MOUNT_TYPE	URBN_TYPE	COAST_TYPE	FID	geometry
0	BG423	BG423	3	BG	Pazardzhik	Пазарджик	3.0	2	3	BG423	POLYGON ((24.42101 42.55306, 24.41032 42.46950...
1	BG424	BG424	3	BG	Smolyan	Смолян	3.0	3	3	BG424	POLYGON ((25.07422 41.79348, 25.05851 41.75177...
2	BG425	BG425	3	BG	Kardzhali	Кърджали	3.0	3	3	BG425	POLYGON ((25.94863 41.32034, 25.90644 41.30757...

It is good practice to set an index. You can use .set_index() for that:

nuts = nuts.set_index('id')

We can also check out the geometries in the dataset with .geometry:

nuts.geometry

id
BG423    POLYGON ((24.42101 42.55306, 24.41032 42.46950...
BG424    POLYGON ((25.07422 41.79348, 25.05851 41.75177...
BG425    POLYGON ((25.94863 41.32034, 25.90644 41.30757...
CH011    MULTIPOLYGON (((6.86623 46.90929, 6.89621 46.9...
CH012    POLYGON ((8.47767 46.52760, 8.39953 46.48872, ...
                               ...                        
LV       POLYGON ((27.35158 57.51824, 27.54521 57.53444...
ME       POLYGON ((20.06394 43.00682, 20.32958 42.91149...
MK       POLYGON ((22.36021 42.31116, 22.51041 42.15516...
SK0      POLYGON ((19.88393 49.20418, 19.96275 49.23031...
IT       MULTIPOLYGON (((12.47792 46.67984, 12.69064 46...
Name: geometry, Length: 2010, dtype: geometry

With .crs we can check in which coordinate reference system the data is given:

nuts.crs

<Geographic 2D CRS: EPSG:4326>
Name: WGS 84
Axis Info [ellipsoidal]:
- Lat[north]: Geodetic latitude (degree)
- Lon[east]: Geodetic longitude (degree)
Area of Use:
- name: World.
- bounds: (-180.0, -90.0, 180.0, 90.0)
Datum: World Geodetic System 1984 ensemble
- Ellipsoid: WGS 84
- Prime Meridian: Greenwich

nuts.total_bounds

array([-63.08825, -21.38917,  55.83616,  80.76427])

Let’s filter by NUTS-1 level…

nuts1 = nuts.query("LEVL_CODE == 1")

… and explore what kind of geometries we have in the dataset …

nuts1.explore()

Make this Notebook Trusted to load map: File -> Trust Notebook

To write a GeoDataFrame back to file use GeoDataFrame.to_file(). The file format is inferred from the file ending.

nuts1.to_file("NUTS1.geojson")

Calculating the areas and buffers

The first thing we need to do to calculate area or buffers is to reproject the GeoDataFrame to an equal-area projection (here: EPSG:3035):

nuts1 = nuts1.to_crs(3035)

The area can be accessed via .area and is given in m² (after projection). Let’s convert to km²:

area = nuts1.area / 1e6
area

id
AT1    23545.286205
AT2    25894.953057
EL4    17388.679384
EE0    45315.713593
EL3     3799.676547
           ...     
PL7    29846.398582
PL8    63217.536546
PL9    35563.812826
RO2    72545.290405
SK0    49008.115415
Length: 125, dtype: float64

nuts1.explore(column=area, vmax=1e5)

Make this Notebook Trusted to load map: File -> Trust Notebook

We can also build a buffer of 1km around each geometry using .buffer():

nuts1.buffer(1000).explore()

Make this Notebook Trusted to load map: File -> Trust Notebook

Joining spatial datasets

Multiple GeoDataFrames can be combined via spatial joins.

Observations from two datasets are combined with the .sjoin() function based on their spatial relationship to one another (e.g. whether they are intersecting or overlapping). You can read more about the specific options here.

Let’s first reproject the gdf object to the same CRS as nuts1:

gdf = gdf.to_crs(3035)

Then, let’s have a look at both datasets at once. We want to find out which points (representing power plants) lie within which shape (representing NUTS regions).

fig = plt.figure(figsize=(7,7))

ax = plt.axes(projection=ccrs.epsg(3035))

nuts1.plot(
    ax=ax,
    edgecolor='black',
    facecolor='lightgrey'
)

gdf.to_crs(3035).plot(
    ax=ax,
    column='Fueltype',
    markersize=gdf.Capacity/20,
    legend=True
)

ax.set_extent([5, 19, 47, 55])

We can now apply the .sjoin function to look for which power plants lie within which NUTS1 region. By default, sjoin looks for intersections and keeps the geometries of the left GeoDataFrame.

joined = gdf.sjoin(nuts1)

If we look at this new GeoDataFrame, we now have additional columns from the NUTS1 data:

joined.head(3)

	Name	Fueltype	Technology	Set	Country	Capacity	Efficiency	DateIn	DateRetrofit	DateOut	...	index_right	NUTS_ID	LEVL_CODE	CNTR_CODE	NAME_LATN	NUTS_NAME	MOUNT_TYPE	URBN_TYPE	COAST_TYPE	FID
id
0	Emsland	Nuclear	Steam Turbine	CHP	Germany	1336.000000	0.33000	1988.0	1988.0	2022.0	...	DE9	DE9	1	DE	Niedersachsen	Niedersachsen	0.0	0	0	DE9
399	Wolfsburg West	Hard Coal	CCGT	CHP	Germany	282.112709	0.38582	1985.0	2018.0	2022.0	...	DE9	DE9	1	DE	Niedersachsen	Niedersachsen	0.0	0	0	DE9
401	Buschhaus	Lignite	Steam Turbine	PP	Germany	352.000000	0.36400	1985.0	2002.0	2020.0	...	DE9	DE9	1	DE	Niedersachsen	Niedersachsen	0.0	0	0	DE9

3 rows × 29 columns

We can now use these new columns to group the capacities (and convert to a suitable unit):

cap = joined.groupby("NUTS_ID").Capacity.sum() / 1000 # GW

Let’s quickly check if all NUTS1 regions have power plants:

nuts1.index.difference(cap.index)

Index(['CY0', 'ES7', 'FI2', 'FRY', 'IS0', 'LI0', 'MK0', 'MT0', 'PT2', 'PT3',
       'TR1', 'TR2', 'TR3', 'TR4', 'TR5', 'TR6', 'TR7', 'TR8', 'TR9', 'TRA',
       'TRB', 'TRC'],
      dtype='object')

This is not the case. Then it is good practice to reindex the series to include all NUTS1 regions, even if this leads to some NaN values.

cap = cap.reindex(nuts1.index)

cap

id
AT1    4.387100
AT2    3.782500
EL4    0.427967
EE0    2.014000
EL3    1.310142
         ...   
PL7    6.561524
PL8    1.146551
PL9    5.896788
RO2    2.462566
SK0    5.697220
Name: Capacity, Length: 125, dtype: float64

Finally, we can plot the total generation capacity per NUTS1 region on a map.

nuts1.plot(figsize=(7,7), column=cap, legend=True)

<Axes: >

This concludes the geopandas and cartopy tutorial.

Exercises

Task 1: Recreate the figure above (i.e. generation capacity per NUTS1 region)

• using 3 different cartopy projections of your choice.

• such that the capacities are normalised by the area of the NUTS1 region (unit: MW/km²).

• such that it only shows the hard coal power plant capacities.

• such that it only shows NUTS1 shapes for Germany (and Denmark).

• but with NUTS2 regions instead of NUTS1 regions.

• (advanced) such that it shows the capacity-weighted average age of the power plants instead of the capacities. Where are the oldest generators? Can you think of a reason why?

Which German NUTS2 region has the highest conventional generation capacity? Can you think of a reason why?

Three projections of choice:

for crs in [ccrs.EqualEarth(), ccrs.AlbersEqualArea(), ccrs.Orthographic()]:
    fig = plt.figure()
    ax = plt.axes(projection=crs)
    nuts1.to_crs(crs.proj4_init).plot(ax=ax, column=cap, legend=True)

Normalised by area in MW/km²:

nuts1.plot(figsize=(7,7), column=cap/(nuts1.area/1e9), legend=True)

<Axes: >

Only hard coal generators:

hard_coal_cap = (
    joined
    .query("Fueltype == 'Hard Coal'")
    .groupby("NUTS_ID").Capacity.sum()
    .reindex(nuts1.index)
    .div(1e3)
)
nuts1.plot(
    figsize=(7,7),
    column=hard_coal_cap,
    legend=True,
    missing_kwds=dict(color='lightgrey')
)
plt.ylim(1e6, 6e6)
plt.xlim(2e6, 6e6);

Only Germany:

subregion = nuts1.query("CNTR_CODE == 'DE'")
subregion.plot(column=cap.reindex(subregion.index), legend=True)

<Axes: >

Only Germany and Denmark:

countries = ['DE', 'DK']
subregion = nuts1.query("CNTR_CODE in @countries") # alternative A
subregion = nuts1.loc[nuts1.CNTR_CODE.isin(countries)] # alternative B
subregion.plot(column=cap.reindex(subregion.index), legend=True)

<Axes: >

In NUTS2 rather than NUTS1:

nuts2 = nuts.query("LEVL_CODE == 2").to_crs(3035)
joined2 = gdf.sjoin(nuts2)
cap2 = joined2.groupby("NUTS_ID").Capacity.sum().reindex(nuts2.index).div(1000) # GW
nuts2.plot(figsize=(7,7), column=cap2, legend=True)

<Axes: >

NUTS2 region with highest capacity:

cap2.sort_values(ascending=False).head(1) # alternative A

id
FRK2    22.54195
Name: Capacity, dtype: float64

cap2.idxmax() # alternative B

'FRK2'

Capacity-weighted average age:

def calculate_weighted_age(x):
    """
    Formula:
    \bar{x} = \frac{ \sum\limits_{i=1}^n w_i x_i}{\sum\limits_{i=1}^n w_i}
    """
    
    numerator = (x.Capacity * x.DateIn).sum()
    
    # only build sum of capacities where a date is given
    denominator = x.Capacity.where(~x.DateIn.isna()).sum()
    
    return numerator / denominator

age = joined.groupby("NUTS_ID").apply(calculate_weighted_age).reindex(nuts1.index)

nuts1.plot(figsize=(7,7), column=age, legend=True)

<Axes: >

Task 2: Load the following .gpkg file containing the Danish Natura2000 natural protection areas as GeoDataFrame:

https://tubcloud.tu-berlin.de/s/mEpdmgBtmMbyjAr/download/Natura2000_end2021-DK.gpkg

• How many individual natural protection areas are there?

• What is the name of the largest natural protection area in Denmark?

• What is the total protection area in square kilometers (make sure you don’t double-count overlapping areas by running the .dissolve() function).

• (advanced) Use set operations with the .overlay() function and the NUTS1 regions GeoDataFrame to identify what share of Danish natural protection areas is at sea.

url = "https://tubcloud.tu-berlin.de/s/mEpdmgBtmMbyjAr/download/Natura2000_end2021-DK.gpkg"

natura = gpd.read_file(url)

natura.plot(alpha=0.5)

<Axes: >

natura.crs

<Derived Projected CRS: EPSG:3035>
Name: ETRS89-extended / LAEA Europe
Axis Info [cartesian]:
- Y[north]: Northing (metre)
- X[east]: Easting (metre)
Area of Use:
- name: Europe - European Union (EU) countries and candidates. Europe - onshore and offshore: Albania; Andorra; Austria; Belgium; Bosnia and Herzegovina; Bulgaria; Croatia; Cyprus; Czechia; Denmark; Estonia; Faroe Islands; Finland; France; Germany; Gibraltar; Greece; Hungary; Iceland; Ireland; Italy; Kosovo; Latvia; Liechtenstein; Lithuania; Luxembourg; Malta; Monaco; Montenegro; Netherlands; North Macedonia; Norway including Svalbard and Jan Mayen; Poland; Portugal including Madeira and Azores; Romania; San Marino; Serbia; Slovakia; Slovenia; Spain including Canary Islands; Sweden; Switzerland; Turkey; United Kingdom (UK) including Channel Islands and Isle of Man; Vatican City State.
- bounds: (-35.58, 24.6, 44.83, 84.73)
Coordinate Operation:
- name: Europe Equal Area 2001
- method: Lambert Azimuthal Equal Area
Datum: European Terrestrial Reference System 1989 ensemble
- Ellipsoid: GRS 1980
- Prime Meridian: Greenwich

How many:

len(natura)

350

Name of largest protection area:

i = natura.area.idxmax()
i

309

natura.loc[i]

SITECODE                                              DK00FX112
SITENAME                               Skagens Gren og Skagerak
MS                                                           DK
SITETYPE                                                      B
INSPIRE_ID                                       Dk.nst.ps.SAC1
geometry      POLYGON ((4323183.452199999 3832152.5764000034...
Name: 309, dtype: object

Total area in km²:

total = natura.dissolve()

total.plot(alpha=0.5)

<Axes: >

total.area / 1e6 # sqkm

0    22646.375701
dtype: float64

Share of natural protection area offshore:

onshore_natura = gpd.overlay(nuts1.dissolve(), total)

onshore_natura.explore()

Make this Notebook Trusted to load map: File -> Trust Notebook

1 - onshore_natura.area[0] / total.area[0]

0.8133701227617232